Small Batch Silicon Wafering

Transcription

1 Small Batch Silicon Wafering Stephen Preece PhD 3 rd Annual c-si PVMC Workshop Marriot Marquis, San Francisco July 8 th

2 2 Process Research Products Test Program Wire Saw Aquaslice (Coolant) Setting Wafering Parameters Ingots and Wire Results Conclusions Near Term Challenges Long Term Challenges

3 Process Research Products 1013 Whitehead Road Extension, Trenton, NJ 08638, USA Tel: Formed as a private company in 1958 (Princeton, New Jersey) Acquired by Chessco Industries (Connecticut) in 1970 President: Jeffrey Radler Privately held company Manufacturer of specialty processing chemicals for: Electronics and solar wafers (silicon) Optoelectronics substrates (sapphire, silicon carbide) Automotive industry (bearing balls, differential gears) Cleaning electronics and metal components ISO 9001:2008 registered, REACH and RoHS compliant Member of PVMC since April

4 Overview Produce small batches of monocrystalline silicon wafers with an R&D scale diamond wire saw. Test strategy: Standardize slicing conditions and coolant Vary ingot type Vary wire type Compare results for each wire/ingot combination: Wire wear Wafer shape 4

5 R&D Scale Diamond Wire Saw Selected Features: Compact size 1.56 m high 5 km spool capacity 12 liter coolant tank Max. 15 wires at 0.35 mm pitch Max. wire speed 700 m/min Operates in up cut mode 5

6 Work Box Detail Photo courtesy of GTI Technologies, Shelton, CT 6

7 PRP own brand diamond wire coolant for bricking & slicing silicon Key features: Extends wire life through enhanced wetting, lubrication and soil rejection properties. Wide operating temperature range. Minimal foam, odor and environmental impact. Non-corrosive to equipment; safe to handle; easy to clean up. Compatible with mechanical filtration systems. Use concentration: 2% to 8% dilution in water. Manufactured in the USA. 7

8 Three saw parameters examined in detail Variable Range Preferred Value Notes Work Table Speed mm/min 0.6 mm/min Affects cut time, wire bow, wire marks Wire Supply Rate m/min 2.5 m/min Significantly affects wire break rate # of Wires in Web Wire breaks increased with # of wires in web; max. # of wires limited by wire supply. Most effective saw configuration: Relatively high table speed, high fresh wire supply rate and fewer wires in web. Wire supply rate had the most significant effect on frequency of wire breaks. # of wires also affected incidences of wire breakage. 8

9 Observations/Resolutions General findings: Occurred at different points in wire winding cycle, although mostly at the point where wire changed direction. Frequency of breaks related to fresh wire supply rate and number of wires in web. Not always due to excessive wire wear. Condition of work rollers & pulleys a major factor. Resolutions: Increase fresh wire supply rate. Reduce number of wires in web. Inspect wear parts before every test. Replace wear parts every 30 hours. 9

10 Wire breakage rate decreased with increasing wire supply % Success 10 53% Success Number of Tests % Success Pass Fail Fresh Wire Supply (m/min) 10

11 Standard Conditions Wire speed 600 m/min Number of wires in web 6 Work roller pitch 0.35 mm Work Table speed (up cut) 0.6 mm/min Ingot dimensions 156 mm square Cut time 260 min Rocking angle 0 degrees Rocking speed 0 degrees/min Traverser pitch 0.3 mm Wire spool acceleration 4 s Wire spool constant speed 22 s Wire spool deceleration 4 s Fresh wire supply 2.5 m/min Wire type Electroplated diamond Manufacturer(s) Various Wire tension 20 N Coolant Aquaslice, 2% Coolant tank volume 12 L Coolant flow rate 7 L/min 11

12 Two ingots and three different wires evaluated Ingot Type Dimensions (mm) Geometry A Cz P/Boron <100> 156 x 156 x 100 Pseudo-square, rounded corners, surface ground B MCz P/Boron <100> 156 x 156 x 100 Square, chamfered edges, surface ground Wire Type Core Diameter (microns) Diamond Size (Microns) Typical Diameter (Microns) 1 EP Ni EP Ni EP Ni Aim to produce at least 10 wafers per wire/ingot combination. Two cuts per combination. 12

13 Data averaged over several tests with 6 wires, 2.5 m/min wire supply Wire wear defined as % decrease in wire diameter after use. Measured at 50 m intervals before and after each use. Baseline wear = 0.3%. Total Wear (%) Ingot Wire 1 Wire 2 Wire 3 A B

14 Wire supply and number of wires in web are related to overall wear Normalized Wire Wear (%) Ingot A Ingot B 2 6 Wires 1.0 m/min 6 Wires 2.5 m/min 1 Wire 1.0 m/min Wire : Silicon Contact Ratio 14

15 Average values (Microns) for wafers cut at 2.5 m/min wire supply Wire/Ingot 1/A 1/B 2/A 2/B 3/A 3/B Specification # Wafers * Thickness ± 20 TTV Bow Warp n/a Consistent thickness regardless of wire or ingot. TTV and bow values generally consistent and low. Warp showed most variation. * Will produce more wafers 15

16 Average values (microns) for wire 1, ingot A 1.0 m/min 2.5 m/min Specification # Wafers Thickness ± 20 TTV Bow Warp n/a Increasing wire supply gave wafers with better shape. TTV and bow values approximately halved at 2.5 m/min. Extra fresh wire increases wire to silicon contact ratio for more efficient slicing. 16

17 Small batches of silicon wafers were produced successfully. Wire wear had a major impact on wafering success rate. Relatively high wire supply rates were required as a consequence. Wafer shape compared well with commercial specifications. Wire breaks were the main issue affecting productivity. Minimized by increasing wire supply & reducing # of wires in web. Keeping wire path clear of obstructions is important. Maintenance / replacement of wear parts is essential. Type of monocrystalline silicon influenced ease of wafering. Wires used to slice ingot A (Cz) were less worn than for ingot B (MCz). Wafers from ingots A and B generally had similar thickness, TTV, warp and bow for all wires tested. Wafer shape improved with increased wire feed rate. 17